Time-invariant covariates are variables that do not change over time and are used in statistical models to account for consistent individual differences. They help in controlling for unobserved heterogeneity, allowing for more accurate estimation of the effects of time-varying covariates on an outcome variable.